Instructions to use VGribAI/Tensor-1.5-small with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use VGribAI/Tensor-1.5-small with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="VGribAI/Tensor-1.5-small")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("VGribAI/Tensor-1.5-small", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use VGribAI/Tensor-1.5-small with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf VGribAI/Tensor-1.5-small:F16
# Run inference directly in the terminal:
llama cli -hf VGribAI/Tensor-1.5-small:F16

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf VGribAI/Tensor-1.5-small:F16
# Run inference directly in the terminal:
llama cli -hf VGribAI/Tensor-1.5-small:F16

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf VGribAI/Tensor-1.5-small:F16
# Run inference directly in the terminal:
./build/bin/llama cli -hf VGribAI/Tensor-1.5-small:F16

Use Docker

docker model run hf.co/VGribAI/Tensor-1.5-small:F16

Llama new
LM Studio
Jan

vLLM

How to use VGribAI/Tensor-1.5-small with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "VGribAI/Tensor-1.5-small"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "VGribAI/Tensor-1.5-small",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/VGribAI/Tensor-1.5-small:F16

SGLang

How to use VGribAI/Tensor-1.5-small with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "VGribAI/Tensor-1.5-small" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "VGribAI/Tensor-1.5-small",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "VGribAI/Tensor-1.5-small" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "VGribAI/Tensor-1.5-small",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use VGribAI/Tensor-1.5-small with Ollama:
```
ollama run hf.co/VGribAI/Tensor-1.5-small:F16
```
Unsloth Desktop
Docker Model Runner
How to use VGribAI/Tensor-1.5-small with Docker Model Runner:
```
docker model run hf.co/VGribAI/Tensor-1.5-small:F16
```

Lemonade

How to use VGribAI/Tensor-1.5-small with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull VGribAI/Tensor-1.5-small:F16

Run and chat with the model

lemonade run user.Tensor-1.5-small-F16

List all available models

lemonade list

Atomic Chat

Tensor-1.5-small / README.md

VGribAI

Update README.md

81ed35c verified 4 days ago

preview code

Raw History Blame Contribute Delete

2.51 kB

---
library_name: transformers
pipeline_tag: text-generation
license: apache-2.0
language:
- en
tags:
- text-generation
- conversational
- qwen2
- from-scratch
- english
---
# Tensor 1.5

Tensor 1.5 is a compact conversational language model built from scratch by **VGribAI**.

It opens the Tensor line: the first model whose name carries no parameter size, the first one
trained exclusively in English, and a complete assistant rather than a base checkpoint, since it
has gone through supervised fine-tuning.

![bencmarks](bench.jpg)

## What makes it different

- **No size in the name.** Tensor 1.5 starts a naming scheme where the version identifies the
  model, not the parameter count.
- **English only.** The tokenizer, the pretraining and the fine-tuning are all English. The model
  is not expected to follow other languages.
- **A finished assistant.** Tensor 1.5 follows the chat format, keeps the assistant role and stops
  on its own turn end. It is a model you talk to, not a text continuation base.

## Architecture

| | |
|---|---|
| Decoder | Qwen2 style (12 layers, hidden 640, GQA 8 query / 2 KV heads, SwiGLU) |
| Parameters | 50.26M, tied embeddings |
| Vocabulary | 4000 byte-level BPE, trained for this model |
| Context | 512 tokens |
| Default answer length | 64 tokens |
| Weights | fp32 safetensors, plus a f16 GGUF build |

## Intended use

Short answers, quick summaries, simple rewrites and light chit-chat on weak hardware: CPUs,
old laptops, single-board computers and phones.

## Usage

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
tok = AutoTokenizer.from_pretrained("tensor-1.5")
model = AutoModelForCausalLM.from_pretrained("tensor-1.5")
msgs = [{"role": "user", "content": "Who are you?"}]
ids = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt")
out = model.generate(ids, max_new_tokens=64, eos_token_id=[0, 2], pad_token_id=2)
print(tok.decode(out[0, ids.shape[1]:], skip_special_tokens=True))
```

## Chat format

```
<|im_start|>system
You are Tensor, a small assistant made by VGribAI.<|im_end|>
<|im_start|>user
Hello!<|im_end|>
<|im_start|>assistant
Hello! How can I help you?<|im_end|>
```

Stop tokens: `<|im_end|>` (id 2) and `<|eos|>` (id 0).

## Limits

- English only.
- 512 tokens of context, 64 tokens per answer.
- A very small model: fluent phrasing, but no reliable factual knowledge, arithmetic or long
  reasoning. Treat answers as text, not as facts.
- No memory beyond the current context window.

	---
	library_name: transformers
	pipeline_tag: text-generation
	license: apache-2.0
	language:
	- en
	tags:
	- text-generation
	- conversational
	- qwen2
	- from-scratch
	- english
	---

	# Tensor 1.5

	Tensor 1.5 is a compact conversational language model built from scratch by VGribAI.

	It opens the Tensor line: the first model whose name carries no parameter size, the first one
	trained exclusively in English, and a complete assistant rather than a base checkpoint, since it
	has gone through supervised fine-tuning.

	![bencmarks](bench.jpg)

	## What makes it different

	- No size in the name. Tensor 1.5 starts a naming scheme where the version identifies the
	model, not the parameter count.
	- English only. The tokenizer, the pretraining and the fine-tuning are all English. The model
	is not expected to follow other languages.
	- A finished assistant. Tensor 1.5 follows the chat format, keeps the assistant role and stops
	on its own turn end. It is a model you talk to, not a text continuation base.

	## Architecture

	\| \| \|
	\|---\|---\|
	\| Decoder \| Qwen2 style (12 layers, hidden 640, GQA 8 query / 2 KV heads, SwiGLU) \|
	\| Parameters \| 50.26M, tied embeddings \|
	\| Vocabulary \| 4000 byte-level BPE, trained for this model \|
	\| Context \| 512 tokens \|
	\| Default answer length \| 64 tokens \|
	\| Weights \| fp32 safetensors, plus a f16 GGUF build \|

	## Intended use

	Short answers, quick summaries, simple rewrites and light chit-chat on weak hardware: CPUs,
	old laptops, single-board computers and phones.

	## Usage

	```python
	from transformers import AutoModelForCausalLM, AutoTokenizer

	tok = AutoTokenizer.from_pretrained("tensor-1.5")
	model = AutoModelForCausalLM.from_pretrained("tensor-1.5")

	msgs = [{"role": "user", "content": "Who are you?"}]
	ids = tok.apply_chat_template(msgs, add_generation_prompt=True, return_tensors="pt")
	out = model.generate(ids, max_new_tokens=64, eos_token_id=[0, 2], pad_token_id=2)
	print(tok.decode(out[0, ids.shape[1]:], skip_special_tokens=True))
	```

	## Chat format

	```
	<\|im_start\|>system
	You are Tensor, a small assistant made by VGribAI.<\|im_end\|>
	<\|im_start\|>user
	Hello!<\|im_end\|>
	<\|im_start\|>assistant
	Hello! How can I help you?<\|im_end\|>
	```

	Stop tokens: `<\|im_end\|>` (id 2) and `<\|eos\|>` (id 0).

	## Limits

	- English only.
	- 512 tokens of context, 64 tokens per answer.
	- A very small model: fluent phrasing, but no reliable factual knowledge, arithmetic or long
	reasoning. Treat answers as text, not as facts.
	- No memory beyond the current context window.